Background subtraction and transient timing with Bayesian Blocks

Abstract

Aims: To incorporate background subtraction into the Bayesian Blocks algorithm so that transient events can be timed accurately and precisely even in the presence of a substantial, rapidly variable, background. Methods: We developed several modifications to the algorithm and tested them on a simulated XMM-Newton observation of a bursting and eclipsing object. Results: We found that bursts can be found to good precision for almost all background subtraction methods, but eclipse ingresses and egresses present problems for most methods. We found one method that recovered these events with precision comparable to the interval between individual photons, in which both source and background region photons are combined into a single list and weighted according to the exposure area. We have also found that adjusting the Bayesian Blocks change points nearer to blocks with higher count rate removes a systematic bias towards blocks of low count rate.